I computed log2 fold changes (which is normally distributed) of treated vs untreated. I standardized ((x-mean)/sd) it and computed the p-values based on the assumption of the normal distribution and z-scores. I got some quite high log2 fold changes mainly due to the low counts. Is there a possibility to account for the low-counts (e.g high counts) while computing fold changes or to model the log 2 fold change as another distribution ( like beta distribution) that will count for the low-counts?
(PS. I am aware of DESeq, GFOLD, edgeR. However, I need a simple (maybe not that robust and reliable) method to account for low counts while calculating the p-values)
2 answers
A reasonable model like DESeq, GFOLD or edgeR should not be cursed by low read counts. The calculation is not related to fold change.
A reasonable model like DESeq, GFOLD or edgeR should not be cursed by low read counts. The calculation is not related to fold change.
Log in to answer this question.
-
which data type I can use to bulid machine learning.
written by Qiang •
-
Ranked fold changes for non-replicated data
written by Tom
-
Expression thresholds for enrichment analysis of RNA-Seq data
written by James Cook •
-
Transform log2 fold changes in z-scores
written by gg •
-
RNAseq: What's the accurate way to test whether a set of N genes is differentially expressed?
written by francois •
-
Statistical test to compare expression change in a subgroup of genes versus all genes?
written by biostart
-
log2 reverses mean in two groups
written by sharmi.banerji •
-
Downregulated Genes with GFOLD
written by jth
-
would it be a good idea to normalize log2 fold changes with loess?
written by tonja.r
-
p-values from a distribution of fold-changes
written by tonja.r
